Google Discovers AI Hacking Tool That Exploits Software Without Humans
Google's Threat Intelligence Group (GTIG) has identified an AI hacking tool capable of discovering software vulnerabilities and exploiting them without human intervention. Detailed in a report published on November 16, 2026, the discovery marks a significant development in cybercrime, as attackers increasingly use autonomous AI agents to conduct sophisticated operations. The tool, named exploit agent, was discovered on an exposed staging server containing Russian-language comments in its configuration files.
How Google's AI Hacking Tool Works
Unlike traditional hacking tools that require operators to identify vulnerabilities and execute attacks manually, exploit-agent automates several stages of the process.
The framework continuously monitors popular web applications for software updates. When developers release security patches, it compares the affected code before and after the update to identify the vulnerabilities being fixed.
It then uses multiple large language models (LLMs) to analyse potential weaknesses, generate exploit code and execute attacks automatically.
Google described this as the first time it had observed threat actors using AI frameworks to fully automate vulnerability discovery and exploitation without a human in the loop.
The exposed server also contained stolen victim data and scripts designed for activities such as internal network reconnaissance, password brute-forcing and installing webshells, which allow attackers to maintain access to compromised websites.
Why Russian Comments in the AI Hacking Tool Files Matter
Google found Russian-language comments in the tool's configuration files. The translated instructions referenced API keys needed for automated exploitation and told operators where to enter them.
The comments provide a clue about the tool's operators, but Google did not identify a specific hacking group or establish that the operation was directed by the Russian government.
The report also highlighted separate AI-powered cyber operations involving North Korean threat actors targeting cryptocurrency companies. In those incidents, attackers reportedly used fake recruitment profiles to gain access to organisations and deploy malicious code through developers' coding environments.
The discovery raises concerns for businesses that depend on internet-facing applications, outdated software and third-party development tools. AI-powered attacks can reduce the time between a vulnerability becoming publicly known and attackers exploiting it.
Cybersecurity teams will need to improve automated vulnerability detection, apply security patches quickly and monitor suspicious activity across their networks. AI will increasingly become part of both offensive and defensive cybersecurity operations.
The biggest concern is not simply that hackers are using AI. It is that autonomous AI hacking tools can now connect vulnerability discovery, exploit development and execution into a single process. Businesses can no longer afford to treat software updates as routine maintenance. They are an essential part of cybersecurity defence.